Each data card is classified as one of the three alert levels (Normal, Critical, Anomaly), as indicated by the color on the top left of the card. The Normal and Critical cards belong to clusters that Subject Matter Experts (SME) have identified as normal or criti- cal. The Anomaly cards are event periods in which SparkPredict’s machine learning models are not confident in any pre-existing cluster designation.
The event in each data card depicts the operating state that the cards belong to. Operating states are defined as the individualized macro-operating conditions of the asset. For a combustion turbine, these operating states are Fullspeed, Lowspeed, Startup, and Coastdown.
Cluster represents the behavioral pattern that SparkPredict has detected for each specific data card. Cluster is determined using a variety of machine learning models. The cluster of the card/event can be changed using the drop-down in this field.
Confidence percentage, displayed to the right of the selected cluster, is the confidence SparkPredict has in determining the cluster of a data card.
Top Contributor is the sensor tag that SparkPredict has algorithmically determined as the number one contributor for that data card to be a part of that cluster. This can be used to identify which cluster an anomaly belongs to.
Note: All events within a single cluster will have the same top contributor
START/ END DATE
The start and end time of the data contained within the card.
Time span, displayed by the clock on the bottom left, details the duration of data within a given card. This is useful in particular when looking at anomalous events. Anomaly cards with long time spans are typically of interest to subject matter experts.
Clicking on a card will pull up details. Details within a card allows a user to further analyze a data card and inspect individual contributors. The details section also contains evidence to support why SparkPredict has classified a certain data card into certain clusters (or states). This is explained further in the “Data Card Details” section.